Beach House Center for Recovery
An AMP Case Study
Site Background and Implementation
Beach House Center for Recovery is a leading provider of addiction treatment.
Beach House started with the AMP Automatic plugin. However, they switched to AMP for WordPress, as it allows much more styling, but you have to pay for some of these features. For example, if you pay more, the plugin automatically configures forms and ads for you.
They converted just the blog and learning center articles. Site: returns 477 pages. 224 pages in the Learning Center, and 66 in the blog. That means about 61% of the pages were converted to AMP.
No forms were implemented in AMP due to the additional costs involved. They want to add CTAs to get people to subscribe to the blog. That work is currently in progress.
Implementation took approximately four weeks.
AMP Metrics
Conversions for Beach House come in the form of leads – either a form fill, or a phone call. No specific rankings increases were seen. Their articles are not in the AMP carousel (the site is not in Google News).
The metrics observed for the AMP pages on the site were as follows:
- Pages per session: 82% of what the regular mobile pages get
- Session time: 33% of what the regular mobile pages get
- Bounce rate: 10% more than the regular mobile pages
As you can see, the metrics are not good. Closer examination revealed problems with the AMP pages. First of all, the site uses Google Analytics, and the session stitching fix was not in place (the AMP implementation was done before this fix was available).
The problem that session stitching refers to is that AMP pages are hosted in the Google Content Delivery Network (CDN). When a user clicks to go to another page, they are brought to the main domain (in this case, beachhouserehabcenter.com). Because that is a change of domains, this is seen by the main site as a new visitor, and the analytics on the AMP pages see that as a visit ending.
The above analytics data is from the AMP pages, and without session stitching in place, it’s almost inevitable that the pages per visitor and session time would be quite low, and that the bounce rate would be quite high. In other words, we don’t have a valid measurement.
In addition, there were other problems with the initial AMP implementation. Here is what the Hamburger menu looked like when you clicked on it:
The menu is already open, and looks like it’s presenting the menu choices alphabetically, making it hard to follow, and unlikely to offer the choices a user on an article might be looking for. Sidebar links to related articles were also lost from the implementation, as well as their phone number and an on-site CTA. After some work, the AMP menu was updated to the following:
This should fit much more closely with what a user on the site may be looking for. Work is ongoing to add a CTA and other UI improvements. The session stitching issue fix has also been put in place. These steps are expected to improve the metrics for the pages, but data on that is not yet available.
Implementation Issues
Beach House had a problem because they were using the “CSS important tag,” and AMP doesn’t allow those. There were no WordPress plugin bugs that they had to patch.
No Facebook JavaScript is allowed, only Google code, and as a result the Facebook tracking pixel had to be removed. Facebook Instant articles is now supporting AMP in their new Instant Articles SDK (Facebook's Instant Articles Platform to Support Google AMP).
Initially, they decided not to set up a CTA on the AMP content. Doing so requires a plugin from AMP for WordPress, but they are now moving forward with that purchase and the implementation. Note that some other add-ons that can be purchased include star ratings and forms.
ROI
Overall results were not good, but that was driven by a faulty initial implementation, and in particular, the lack of a valid analytics implementation (because the session stitching fix was not available when the AMP implementation was put into place). Changes have been made to improve those metrics, but new numbers are not available as of the time of publication of this profile. The profile will be updated with new numbers once the new metrics are available.